Finance teams are spending an average of 13 hours per week verifying outputs produced by artificial intelligence tools, according to a new report commissioned by accounting software provider Sage and conducted by International Data Corp. That lost time costs roughly $78,000 annually per senior finance professional - assuming $250,000 in all-in compensation - and erodes roughly 26% of the productivity gains AI is projected to deliver.
The study, titled "The Emerging Economics of AI in Finance," surveyed more than 2,000 senior finance leaders, including 832 based in the U.S. It paints a picture of finance departments caught between ambitious AI adoption goals and the stubborn reality that automated outputs still demand significant human oversight.
The $78,000 cost of double-checking AI
The report defines these lost hours as AI's verification tax. Nearly half of respondents - 48% - said they spend more than 15 hours per week verifying AI outputs. Another 19% spend over 30 hours per week on the task, a threshold the report's authors said marks "the level at which AI creates more work than it saves."
Sage CTO Aaron Harris acknowledged the friction but argued the net result still favors adoption. "A 26% tax still leaves you with 74% return, which is pretty significant," Harris said. "But the focus needs to be on reducing the 26%." He added that it is "very expected" for humans to shift from being doers to reviewers as technology matures.
For finance leaders navigating this shift, AI for Finance Courses often address the verification skills and trust frameworks that underpin effective human-AI collaboration in accounting workflows.
Most finance operations remain manual
Despite the rush to implement AI, fully autonomous finance operations are rare. Just 4% of respondents said they run largely autonomous finance functions, while 62% described their operations as primarily manual or rules-based. The gap between ambition and reality is wide.
Harris pointed to the stakes involved. "When it comes to compliance related work, when it comes to reporting to stakeholders, there's no tolerance for inaccuracy," he said. "That is an instant loss of credibility."
Black box versus glass box tools
The report draws a sharp distinction between black box AI tools - which deliver outputs with little or no explanation - and glass box tools, which provide clear reasoning trails that users can interrogate. Finance leaders show little patience for opacity.
Seven in 10 respondents agreed their organization would intentionally limit AI autonomy without real-time visibility into the agent's logic or a solid post-hoc audit trail. And 71% said they would veto a 99%-accurate AI tool that could not produce a human-readable reasoning trace for every decision. The report noted that "71% agree that the lack of AI transparency fundamentally undermines their ability to fulfill their fiduciary duty to the board."
Harris suggested that fully autonomous actions - such as paying invoices without human approval - remain a bridge too far for most businesses. "That's a level of risk that humans are uncomfortable with," he said. For CFOs and senior finance executives evaluating these systems, the AI Learning Path for CFOs covers how to audit AI tools, assess vendor claims, and set governance standards that match the fiduciary demands of the role.
Why this matters for finance professionals
The 13-hour weekly verification burden is not just a productivity drain - it is a signal about where AI tools still fall short in regulated, high-stakes environments. Finance leaders who ignore the verification tax in their ROI calculations will overestimate near-term savings. Those who invest in glass box tools and build verification competence into their teams will capture more of the 74% net gain Harris describes, while avoiding the credibility damage that comes from an unexplained error passing through unchecked.
Your membership also unlocks:
